Effects of different instructional sets on the construct validity of the NEO-PI-R

The aim of the present study was to compare the effects of faking including changes of the construct validity of the NEO-PI-R evoked by an unspecified fake good and two specifically formulated applicant instructions. Whereas in the fake good instruction all scales were strongly faked, mean faking scores showed a more differentiated pattern within the applicant conditions. In addition, nearly all correlations between NEO-PI-R domains inflated under the instruction to fake good, whereas correlations in the applicant conditions were inflated only for those domains, which were rated as important or unimportant for the respective job. Our results imply that unspecified fake good instructions should not be considered to be representative simulations of faking on personality questionnaires in personnel selection contexts, and applicant instructions should be used instead.